/* * Copyright (c) 2003, 2006 Matteo Frigo * Copyright (c) 2003, 2006 Massachusetts Institute of Technology * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by * the Free Software Foundation; either version 2 of the License, or * (at your option) any later version. * * This program is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * GNU General Public License for more details. * * You should have received a copy of the GNU General Public License * along with this program; if not, write to the Free Software * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA * */ /* $Id: dftw-direct.c,v 1.13 2006-02-13 12:59:06 athena Exp $ */ #include "ct.h" typedef struct { ct_solver super; const ct_desc *desc; int bufferedp; kdftw k; } S; typedef struct { plan_dftw super; kdftw k; INT r, vl; INT s, vs; INT mcount; stride bufstride, ios; const R *tdW; INT mstart, m; twid *td; const S *slv; } P; /************************************************************* Nonbuffered code *************************************************************/ static void apply(const plan *ego_, R *rio, R *iio) { const P *ego = (const P *) ego_; INT i, vl = ego->vl, s = ego->s, vs = ego->vs, mcount = ego->mcount; const R *W = ego->tdW; ASSERT_ALIGNED_DOUBLE; for (i = 0; i < vl; ++i) ego->k(rio + i * vs, iio + i * vs, W, ego->ios, mcount, s); } /************************************************************* Buffered code *************************************************************/ static const R *dobatch(kdftw k, R *rA, R *iA, const R *W, stride ios, INT dist, INT r, INT batchsz, R *buf, stride bufstride) { X(cpy2d_pair_ci)(rA, iA, buf, buf + 1, r, WS(ios, 1), WS(bufstride, 1), batchsz, dist, 2); W = k(buf, buf + 1, W, bufstride, batchsz, 2); X(cpy2d_pair_co)(buf, buf + 1, rA, iA, r, WS(bufstride, 1), WS(ios, 1), batchsz, 2, dist); return W; } /* must be even for SIMD alignment; should not be 2^k to avoid associativity conflicts */ static INT compute_batchsize(INT radix) { /* round up to multiple of 4 */ radix += 3; radix &= -4; return (radix + 2); } static void apply_buf(const plan *ego_, R *rio, R *iio) { const P *ego = (const P *) ego_; INT i, j, mcount = ego->mcount, vl = ego->vl, r = ego->r; INT batchsz = compute_batchsize(r); R *buf; STACK_MALLOC(R *, buf, r * batchsz * 2 * sizeof(R)); for (i = 0; i < vl; ++i) { R *rA = rio + i * ego->vs, *iA = iio + i * ego->vs; const R *W = ego->tdW; for (j = 0; j < mcount - batchsz; j += batchsz) { W = dobatch(ego->k, rA, iA, W, ego->ios, ego->s, ego->r, batchsz, buf, ego->bufstride); rA += ego->s * batchsz; iA += ego->s * batchsz; } dobatch(ego->k, rA, iA, W, ego->ios, ego->s, ego->r, mcount - j, buf, ego->bufstride); } STACK_FREE(buf); } /************************************************************* common code *************************************************************/ static void awake(plan *ego_, enum wakefulness wakefulness) { P *ego = (P *) ego_; X(twiddle_awake)(wakefulness, &ego->td, ego->slv->desc->tw, ego->r * ego->m, ego->r, ego->m); ego->tdW = X(twiddle_shift)(ego->td, ego->mstart); } static void destroy(plan *ego_) { P *ego = (P *) ego_; X(stride_destroy)(ego->bufstride); X(stride_destroy)(ego->ios); } static void print(const plan *ego_, printer *p) { const P *ego = (const P *) ego_; const S *slv = ego->slv; const ct_desc *e = slv->desc; if (slv->bufferedp) p->print(p, "(dftw-directbuf/%D-%D/%D%v \"%s\")", compute_batchsize(ego->r), ego->r, X(twiddle_length)(ego->r, e->tw), ego->vl, e->nam); else p->print(p, "(dftw-direct-%D/%D%v \"%s\")", ego->r, X(twiddle_length)(ego->r, e->tw), ego->vl, e->nam); } static int applicable0(const S *ego, int dec, INT r, INT m, INT mb, INT me, INT s, INT vl, INT vs, R *rio, R *iio, const planner *plnr) { const ct_desc *e = ego->desc; UNUSED(vl); return ( 1 && dec == ego->super.dec && r == e->radix /* check for alignment/vector length restrictions */ && (e->genus->okp(e, rio, iio, m * s, 0, m, mb, me, s, plnr)) && (e->genus->okp(e, rio + vs, iio + vs, m * s, 0, m, mb, me, s, plnr)) ); } static int applicable0_buf(const S *ego, int dec, INT r, INT m, INT mb, INT me, INT s, INT vl, INT vs, R *rio, R *iio, const planner *plnr) { const ct_desc *e = ego->desc; INT batchsz; UNUSED(vl); UNUSED(s); UNUSED(vs); UNUSED(rio); UNUSED(iio); return ( 1 && dec == ego->super.dec && r == e->radix /* check for alignment/vector length restrictions, both for batchsize and for the remainder */ && (batchsz = compute_batchsize(r), 1) && (e->genus->okp(e, 0, ((const R *)0) + 1, 2 * batchsz, 0, m, mb, mb + batchsz, 2, plnr)) && (e->genus->okp(e, 0, ((const R *)0) + 1, 2 * batchsz, 0, m, mb, me, 2, plnr)) ); } static int applicable(const S *ego, int dec, INT r, INT m, INT mb, INT me, INT s, INT vl, INT vs, R *rio, R *iio, const planner *plnr) { if (ego->bufferedp) { if (!applicable0_buf(ego, dec, r, m, mb, me, s, vl, vs, rio, iio, plnr)) return 0; } else { if (!applicable0(ego, dec, r, m, mb, me, s, vl, vs, rio, iio, plnr)) return 0; } if (NO_UGLYP(plnr) && X(ct_uglyp)((ego->bufferedp? (INT)512 : (INT)16), m * r, r)) return 0; if (m * r > 262144 && NO_FIXED_RADIX_LARGE_NP(plnr)) return 0; return 1; } static plan *mkcldw(const ct_solver *ego_, int dec, INT r, INT m, INT s, INT vl, INT vs, INT mstart, INT mcount, R *rio, R *iio, planner *plnr) { const S *ego = (const S *) ego_; P *pln; const ct_desc *e = ego->desc; static const plan_adt padt = { 0, awake, print, destroy }; A(mstart >= 0 && mstart + mcount <= m); if (!applicable(ego, dec, r, m, mstart, mstart + mcount, s, vl, vs, rio, iio, plnr)) return (plan *)0; pln = MKPLAN_DFTW(P, &padt, ego->bufferedp ? apply_buf : apply); pln->k = ego->k; pln->ios = X(mkstride)(r, m * s); pln->td = 0; pln->tdW = 0; pln->r = r; pln->m = m; pln->s = s; pln->vl = vl; pln->vs = vs; pln->mstart = mstart; pln->mcount = mcount; pln->slv = ego; pln->bufstride = X(mkstride)(r, 2 * compute_batchsize(r)); X(ops_zero)(&pln->super.super.ops); X(ops_madd2)(vl * (mcount/e->genus->vl), &e->ops, &pln->super.super.ops); if (ego->bufferedp) { /* 8 load/stores * N * VL */ pln->super.super.ops.other += 8 * r * mcount * vl; } pln->super.super.could_prune_now_p = (!ego->bufferedp && r >= 5 && r < 64 && m >= r); return &(pln->super.super); } static void regone(planner *plnr, kdftw codelet, const ct_desc *desc, int dec, int bufferedp) { S *slv = (S *)X(mksolver_ct)(sizeof(S), desc->radix, dec, mkcldw); slv->k = codelet; slv->desc = desc; slv->bufferedp = bufferedp; REGISTER_SOLVER(plnr, &(slv->super.super)); if (X(mksolver_ct_hook)) { slv = (S *)X(mksolver_ct_hook)(sizeof(S), desc->radix, dec, mkcldw); slv->k = codelet; slv->desc = desc; slv->bufferedp = bufferedp; REGISTER_SOLVER(plnr, &(slv->super.super)); } } void X(regsolver_ct_directw)(planner *plnr, kdftw codelet, const ct_desc *desc, int dec) { regone(plnr, codelet, desc, dec, /* bufferedp */ 0); regone(plnr, codelet, desc, dec, /* bufferedp */ 1); }